Abstract: n order to provide additional evidence of the phylogenetic relationships between genera of Rosaceae, it is neccessary to study and analyse the characters in primitive genera of the family. In this paper, leaf epidermis of the primitive genera of Rosaceae, including Kagneckia, Quillaja, Vauquelinia, Lindleya, Exochorda and Lyonothamus of Trib. Quillajeae circumscribed by Hutchinson (1964), has been anatomically studied and description of epidermal structure is presented. Several types of stomatal apparatuses have been observed in the genera investigated and considered as having systematic significance at generic level. (I) On the basis of the foliar epidermal characters, the following key to the genera is presented: 1. Stomatal apparatuses polytypic, of 2-3 types 2. Polytypic stomatal apparatuses incruding 3 types, anomocytic, actinocytic and staurocytic. 3. Polytypic stomatal apparatuses only of 2 types, anomocytic and actinocytic, with the former dominant and surrounding the latter ...... 1. Kagneckia Ruiz & Pav. 3. Polytypic stomatal apparatuses of 3 types, anomocytic, actinocytic and staurocytic 4. Stomatal apparatuses sparse, with anomocytic and actinocytic ones dominant. ....... ..................................................................... 3. Exochorda Lindl. 4. Stomatal apparatuses dense, with staurocytic type dominant. ......................... ........................................................... 4. Lindleya H. B. & K. Nov. 2. Polytypic stomatal apparatuses of 2 types, paracytic and cyclocytic, subsidiary cells narrow and distinguishable from other epidermal cells. ................ 2. Quillaja Molina 1. Stomatal apparatus monotypic, almost only actinocytic. 5. Guard cells surrounded radically by 7-9 subsidiary cells equal in diam. ....... ....................................... 5. Vauquelinia Gorrea ex Humb & Bongl. 5. Guard cells sunkes, 19.76-30.70μm deep, adaxiaL wall of elongated subsidiary cells so united transversely and thickened that the whole stomatal apparatus pot-like, stomatal apparatuses aggregated in a mass and having no any ordinary epidermal cells among them... ........................................................................... 6. Lyonothamus A. Gray (II) In the most primitive extant genus, Kagneckia, the stomatal apparatuses are mainly the most primitive anomocytic, but also actinocytic. (III) In another primitive genus, Quillaja, only second to and most related to the genus Kagneckia, there are both paracytic and cyclocytic types of stomatal apparatuses, some subsidiary cells of which are partly overlapped by the guard cells so that they appear very narrow in surface view. Based on the study of the most primitive genus, Kagneckia, and the second primitive genus, Quillaja, it seems that the paracytic type of stomatal apparatus may be derived from the anomocytic one in Rosaceae. (IV) The stomatal apparatus, in the genus Vauquelinia are actinocytic and undoubtedly derived from that in the more primitive rosoids. (V) The pot-like stomatal apparatuses in the genus Lyonothamus are very unique and_ apparently more specialized. Morphologivally, in the genus, there are opposite, fern-like leaves and the pistil consists of only 2. carpels It may be a very isolated relict along the coast of California in N. America. (VI) In the genus Exochorda, although anomocytic and actinocytic stomatal apparatuses are dominant, staurocytic type also occurs. Morphologically, the genus is more advanced than Kagneckia and Quillaja. In the genus Lindleya, there are also anomocytic and actinocytic types of stomatal, apparatuses, though staurocytic type is dominant. The genus Lindleya is also morphologically more advanced than the most primitive genera of Rosaceae mentioned above. The studies support the inferrence from cytology (Goldblatt 1976) that the tribe Quillajeae in Rosaceae is an unnatural alliance. Acknowledgement The author wishes to express his thanks to Dr. Stevens, who was the supervisor of the Harvard University Herbaria, for allowing me to collect the materials investigated during my stay in the herbaria from 1982 to 1984.